This study presents the technology to combine the remote sensing image of SPOT and FORMOSAT-2 satellite image by Fuzzy theory and Back Propagation Network (BPN). This method adopt three experience identify factors of NDVI, shape, and color to establish the membership function. The fuzzy results show that the successful rate of identification was about 87 percent. The BPN results show that the successful rate of identification was about 91 percent. Therefore the result proves that the three indexes are the best choice in extracting of change area from image, because the error is the lowest among the two methods in this study.